A method for predicting eye movement in a head mounted display (HMD). The method including tracking movement of an eye of a user with a gaze tracking system disposed in the HMD at a plurality of sample points. The method including determining velocity of the movement based on the movement of the eye. The method including determining that the eye of the user is in a saccade upon the velocity reaching a threshold velocity. The method including predicting a landing point on the display of the HMD corresponding to a direction of the eye for the saccade.
Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
3. The method of claim 2, wherein the recurrent neural network implementing the classifier and modeled saccade is configured as a long short term memory (LSTM) network or a fully connected multilayer perceptron network.
5. The method of claim 1, further comprising: training a deep learning engine to generate the plurality of modeled velocity graphs based on training data of measured saccades of a plurality of test subjects.
12. The computer system of claim 8, the method further comprising: training a deep learning engine to generate the plurality of modeled velocity graphs based on training data of measured saccades of a plurality of test subjects.
17. The non-transitory computer-readable medium of claim 16, wherein the recurrent neural network implementing the classifier and modeled saccade is configured as a long short term memory (LSTM) network or a fully connected multilayer perceptron network.
The invention relates to a computer-implemented system for analyzing eye movement data, specifically saccades, to classify or predict user behavior or cognitive states. The system addresses the challenge of accurately modeling and interpreting rapid eye movements (saccades) to infer meaningful patterns, which is difficult due to their transient nature and the complexity of neural processing involved. The solution involves a recurrent neural network (RNN) that functions as a classifier and simultaneously models saccade dynamics. This dual functionality allows the system to process sequential eye movement data while learning temporal dependencies, improving prediction accuracy. The RNN can be configured as either a long short-term memory (LSTM) network, which is well-suited for capturing long-term dependencies in sequential data, or a fully connected multilayer perceptron (MLP) network, which provides flexibility in feature extraction. The system processes input data representing eye movement trajectories, including saccade onset, duration, and direction, to generate outputs such as user intent, attention focus, or cognitive load. The use of an RNN enables real-time or near-real-time analysis, making the system applicable in fields like human-computer interaction, neurodiagnostics, and adaptive user interfaces. The invention improves upon prior methods by integrating saccade modeling directly into the classification process, reducing latency and enhancing interpretability of eye movement data.
19. The non-transitory computer-readable medium of claim 15, further comprising: program instructions for training a deep learning engine to generate the plurality of modeled velocity graphs based on training data of measured saccades of a plurality of test subjects.
The invention relates to eye-tracking technology, specifically systems for modeling and analyzing saccadic eye movements. The problem addressed is the need for accurate, data-driven models of eye movement dynamics to improve applications such as gaze prediction, user interface design, and medical diagnostics. The invention involves a computer-implemented method that generates a plurality of modeled velocity graphs representing saccadic eye movements. These models are derived from training data consisting of measured saccades from multiple test subjects. A deep learning engine is trained on this data to produce the velocity graphs, which can then be used to simulate or analyze eye movement patterns. The system may also include preprocessing steps to normalize or filter the training data, ensuring robustness in the generated models. The deep learning engine is configured to learn the complex relationships between input saccade data and output velocity profiles, enabling high-fidelity simulations of eye movement dynamics. This approach improves upon traditional methods by leveraging large datasets and advanced machine learning techniques to capture individual and population-level variations in saccadic behavior. The resulting models can be applied in fields such as human-computer interaction, vision research, and assistive technologies.
Cooperative Patent Classification codes for this invention. Click any code to explore related patents in that topic.
March 9, 2021
October 18, 2022
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.